This paper presents a simple and fast approach for genome scaffolding, combining constraint modeling and simple graph manipulation. We model the scaffolding problem as an optimization problem on a graph built from a paired-end reads alignment on contigs, then describe an heuristic to solve this problem with the iterative combination of local constraints solving and cycle breaking phases. We tested our approach on a benchmark of various genomes, and compared it with several usual scaffolders. The proposed method is quick, flexible, and provides results comparable to other scaffolders in terms of quality. In addition, contrarily to state of the art approaches that require dedicated servers, it can be run on a basic notebook computer even for large genomes.